Este proyecto presenta una solución de Computer Vision para la detección y clasificación de objetos en imágenes, las cuales son extraídas como frames de vídeos. Utiliza el modelo FastSAM para la detección de objetos, y para la clasificación, emplea embeddings que pueden ser generados mediante dos modelos distintos: CLIP o SigLIP.

Download flickr8k, flickr30k image caption datasets

A minimal, but effective implementation of CLIP (Contrastive Language-Image Pretraining) in PyTorch

Framework for learning multi-domain image embeddings suitable for multi-domain image retrieval at instance-level

Web scraper for Wildberries + simple vectorization/multimodal embedding workflow

Low-latency ONNX and TensorRT based zero-shot classification and detection with contrastive language-image pre-training based prompts

Notes for the Vision Language Model implementation by Umar Jamil

[NeurIPS 2024] AWT: Transferring Vision-Language Models via Augmentation, Weighting, and Transportation

A simple open-sourced SigLIP model finetuned on Genshin Impact's image-text pairs.

Meme search and discovery engine using OpenAI CLIP and Salesforce BLIP

Implementation of PaliGemma

Chitrarth: Bridging Vision and Language for a Billion People

Projects based on SigLIP (Zhai et. al, 2023) and Hugging Face transformers integration 🤗

Leverage SigLIP 2's capabilities using LitServe.

CLIP & SigLIP model training from scratch

SAT-Landforms-Classifier is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to classify satellite images into different landform categories using the SiglipForImageClassification architecture

Gender-Classifier-Mini is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to classify images based on gender using the SiglipForImageClassification architecture.

Fire-Detection-Siglip2 is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to detect fire, smoke, or normal conditions using the SiglipForImageClassification architecture.

Painting-126-DomainNet is an image classification vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for a single-label classification task. It is designed to classify paintings into 126 domain categories using the SiglipForImageClassification architecture

A custom Vision-Language Model (VLM) built from scratch, using SigLip for contrastive learning and a ViT-based encoder to generate meaningful image captions and semantic descriptions.